So I basically have a mediator that I want to test on two other variables. I have the following one-tailed hypothesis I'm testing:

For the sake of simplicity, X is the mediator.

A to B; A to C; A to B via X; A to C via X; and B to C.

A is a 2nd order construct made up of 5 dimensions. A itself as no indicators.

I' using the Repeated Indicator Approach (B) for the 2nd order construct, as suggested here: Article Hierarchical Latent Variable Models in PLS-SEM: Guidelines f...

The trouble I have is, if I keep my dimensions in the model, I can't see the overall indirect effect of A. But I can use the two-stage method (also in the paper above) for it. But, what would be the best way of of doing it?

My best guess right now is to run it with the first way for everything, and then again the second way only to look at the mediation.

Since my model is not calling for any further dependencies with the mediation; I think running them separately will be the best option, but just looking for a second-opinion.

Thanks!